Mojo function

shared_memory_epilogue

shared_memory_epilogue[MMA_M: UInt, data_paths: UInt, num_stages: UInt, stage: UInt, stageN: UInt, c_type: DType, shared_n: UInt, simd_size: UInt, c_smem_upper_layout: Layout, c_smem_lower_layout: Layout, swizzle: Swizzle, compute_lambda_fn: elementwise_compute_lambda_type, num_output_warps: UInt](M: UInt32, N: UInt32, c_col: UInt, c_row: UInt, c_smem_warp_tile_upper: LayoutTensor[c_type, c_smem_upper_layout, MutAnyOrigin, address_space=AddressSpace.SHARED, element_layout=element_layout, layout_int_type=layout_int_type, linear_idx_type=linear_idx_type, masked=masked, alignment=alignment], c_smem_warp_tile_lower: LayoutTensor[c_type, c_smem_lower_layout, MutAnyOrigin, address_space=AddressSpace.SHARED, element_layout=element_layout, layout_int_type=layout_int_type, linear_idx_type=linear_idx_type, masked=masked, alignment=alignment])

Apply element-wise epilogue to non-transposed SMEM tile.

Each warp processes upper (rows 0-15) and lower (rows 16-31) fragments. Uses distribute layout to map SIMD vectors to threads within each warp.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit